[Definitive Guide] The Maximum Breast Size Prompt for AI Images — Conclusions from 30+ Condition Testing

Which prompt maximizes breast size? Cup size notation, cm measurements, weight syntax — many intuitively obvious approaches don’t work. After comparing 30+ conditions in a controlled environment, we identified the descriptions that actually work.

Full experimental data: see Comprehensive Breast Size Prompt Comparison.

Target model: z-image-turbo (6B, photorealism-focused distilled model, CFG=1.0). Results may differ on other models.

Conclusion: Top 3 Maximum Breast Size Prompts

Absolute Maximum: 11-Word Full Stack

NSFW - クリックで表示

seed 1	seed 2	seed 3

The largest result across all 30+ conditions. However, spending 11 words (~15 tokens) purely on size leaves significantly less room for composition, lighting, and outfit. Note also that nipple exposure is more likely.

Efficiency Maximum (Recommended): 5 Words

NSFW - クリックで表示

seed 1	seed 2	seed 3

Achieves near-maximum size from the 11-word stack using less than half the word count. The remaining tokens can be used for composition and lighting, making this the most balanced practical choice.

Quick and Easy: 2 Words

Just add massive tits to the end of your prompt. 2 words get you ★★★★☆ size — the simplest approach.

Why Does Stacking Adjectives Make Them Larger?

The principle behind increasing breast size is the size token volume hypothesis.

The text encoder responds to the total volume of size-meaning tokens in the prompt to determine size. There’s no specific “magic spell” — more size-related adjectives means larger results.

This hypothesis was confirmed with these 3 conditions:

In other words, super strong super huge gigantic breasts is not a magic spell — it just has many size adjectives. You can get the same effect by choosing your own different size adjectives.

Details: Stacking Effect and Size Token Volume Hypothesis

6 Things That Don’t Work

A collection of intuitively plausible approaches that actually fail.

Why Doesn’t Weight Syntax Work?

z-image-turbo operates at CFG=1.0. For unknown reasons, no breast size change from weight syntax was confirmed on this model. Weights of 1.3 through 2.0 all show no visual difference from weight 1.0.

Why Are Metaphors Dangerous?

Writing watermelon-sized breasts causes CLIP to interpret “watermelon” and “breasts” independently. The result: a watermelon appears in the image, and breast size stays the same. holstein breasts produces a cow, and hanging bombs produces an explosion in the background.

The Size Adjective Ladder

A step-by-step view of single-word effectiveness:

Control (no adjective)
  ↓ ★★☆☆☆
large breasts
  ↓ ★★★☆☆ (ceiling here — switching words above huge makes no difference)
huge breasts ≈ gigantic breasts ≈ super huge breasts
  ↓ ★★★★☆ (single adjectives with highest effect)
massive tits ≈ enormous boobs ≈ oversized breasts
  ↓ ★★★★★ (stacking)
oversized massive tits, enormous boobs (5 words)
  ↓ ★★★★★+ (more stacking)
super strong super huge gigantic breasts, massive tits, enormous boobs (11 words)

Notable points:

• tits/boobs tend to produce slightly larger results than breasts, but with higher nipple exposure rates. (Direct comparison with adjectives unified has not been done.)
• oversized is the strongest single adjective — exceeds huge with one word
• Switching words alone has a size ceiling. You need stacking to break through

Details: Step-by-Step Size Adjective Comparison, oversized: The Strongest Adjective

Practical Tips: Hiding Nipples

Making breasts larger increases the chance of nipple exposure. Outfit changes can help suppress exposure.

Note: We tested whether using breasts (instead of tits/boobs) hides nipples, but N=13 all had nipple exposure. Word choice doesn’t hide nipples — only outfit changes can. Even outfit changes don’t guarantee 100% consistency, so generating multiple images and selecting is necessary.

Details: Hiding Nipples with Outfit Changes, breasts Unification Test

How to Use These Prompts

Placing breast size descriptions at the end of the prompt is recommended. Within the token limit of the text encoder, write subject, composition, and lighting first, then add size descriptions last.

Estimated Token Budget

The 5-word efficiency version uses about 10% of the token budget to achieve maximum-class size. The remaining 90% can go to composition, lighting, and outfit — without compromising overall image quality.

Summary

The principle is simple: the more size-meaning adjectives, the larger the breasts. Cup size notation, cm measurements, and weight syntax don’t work.

Lab Director comment: 30 conditions, and the conclusion is “pile on the size words.” I love how clean that is. The 5-word version leaves enough tokens for composition and lighting — practically speaking, that’s the one to use. And the fact that cup size notation does nothing is worth memorizing.

Related Articles

[Verified] Image Generation Prompt Best Practices

Breast Size Prompts That Work — Size Adjectives, Stacking, and Outfit Control